Modeling Correlated Autonomy Failure
For treaty partners. How Castra models cat-style correlation across portfolios that share base models, retrieval indexes, or upstream vendors. Why one drift event can fire across three uncorrelated lines.
Underwriting memos, loss anatomies, and regulatory briefs from Castra Risk. Published quarterly. Read by general counsels, chief risk officers, and treaty underwriters.
The founding memo of the Quarterly. A short note on what Castra writes, why, and what it deliberately does not. Issue No. 001 sets the posture; subsequent issues are technical.
“The discipline is old. The risk is new. The form is, of necessity, both.”
Why prior-art actuarial methods — chain-ladder, Bornhuetter-Ferguson, frequency-severity — fail when the underlying distribution is non-stationary. The three instruments Castra conditions on instead.
The Act's risk tiers, the provider-versus-deployer trap in Article 25, the parallel civil-liability track. What corporate insurance towers must absorb before August 2026, and where affirmative AI coverage sits.
For treaty partners. How Castra models cat-style correlation across portfolios that share base models, retrieval indexes, or upstream vendors. Why one drift event can fire across three uncorrelated lines.
What we ask for, what we accept, and how we treat the streams in pricing. A reference document for chief technology officers preparing for a first AI bind.
One memo per quarter. No promotional sequences. Unsubscribe with one click.